The Ideal Mean Arterial Pressure Target Debate: Heterogeneity Obscures Conclusions*

Both the severity and duration of hypotension are associated with poor outcomes in critically ill patients (1) and during the perioperative period (2). In addition, there is a complex interaction between severity and duration of hypotension. Specifically, a shorter time of exposure to more severe hypotension may be sufficient to increase mortality compared with milder hypotension (1,3). It is interesting to note that this relationship remains valid for systolic and diastolic pressures as well as mean arterial pressure (MAP) (4). In addition, hypotension is strongly associated with occurrence of acute kidney injury and myocardial events (3). Higher pressures have been targeted to prevent some of these negative effects of hypotension. However, increasing MAP target is not without challenges. This usually necessitates the use of more or prolonged durations of vasopressors, which may increase the risk of arrhythmias, organ ischemia, and/or needlessly prolonged ICU stay. In this issue of Critical Care Medicine, D'Amico et al (5) recently conducted a systematic review and meta-analysis of randomized trials investigating lower and higher blood pressure targets in critically ill and surgical patients. The authors included 28 studies (12 in critically ill and 16 in surgical patients), totalizing 15,672 patients, and evaluated the impact of blood pressure targets on all-cause mortality at the longest follow-up available. They used different statistical approaches including frequentist and Bayesian analyses and multiple sensitivity analyses. The authors reported that patients in the lower blood pressure target group had lower mortality (relative risk 0.93; 95% CI, 0.87–0.99; p = 0.03, which corresponded to a 97.4% probability of any increase in mortality with a Bayesian approach). They also observed a lower rate of atrial fibrillation and fewer patients requiring transfusion in lower-pressure target groups. However, there were no differences in the other secondary outcomes, especially in the rates of acute kidney injury. We congratulate the authors on this study, which was an important attempt to sort out the scientific confusion around ideal blood pressure targets in the ICU and during the perioperative period. There was a very robust signal confirmed by a strong statistical analysis. Although observed in heterogeneous populations, multiple sensitivity analyses trying to account for the heterogeneity of these included trials, confirmed the main findings. Indeed, the magnitude and direction of the results were similar in most sensitivity analyses. Of note, there was a huge heterogeneity in the aggregated trials. For example, these trials included patients with varied and different primary insults and conditions, ranging from hip replacement to cardiac arrest and septic shock, submitted to a variable duration of exposure to intervention (from a couple of hours during the operative period to several days in ICU trials). To make things more complicated MAP targets were heterogeneous within similar populations. In two trials of sepsis, the low MAP target in one trial (6) was similar to the higher MAP target in the other trial (7). Although multiple sensitivity analyses attempted to homogenize the comparisons, these could only partially address the issue of heterogeneity. As an example, analyzing separately short vs. long-duration attempts to reach MAP targets could not completely circumvent the issue of mixing hip replacement and bleeding trauma in short duration infusion group and cardiac arrest and septic shock in long-duration group of studies. The sole way to markedly reduce the heterogeneity of the studies would have been to group studies using multiple similar criteria (same duration of exposure to intervention in patients with similar diagnoses and submitted to similar MAP targets). However, this would bring together only 2–3 trials, making the exercise of meta-analysis futile. Clearly, the perioperative group had no difference in mortality, but this is not reassuring. This perioperative group is so heterogeneous that benefit or harm may have been diluted by huge differences in trials. There are important differences in types of surgeries, exposure duration, presence/absence of bleeding, duration of anesthetic agents' exposure, comorbidities, etc. Also in surgical sepsis, there is a huge difference in outcome between surgeries achieving or not achieving adequate source control. Will this study by D'Amico et al (5) modify clinical practice? Maybe not because the bedside interpretation of low and high targets is difficult. In the absence of a true translation of these MAP targets into numerical ranges, the average clinician may feel lost in trying to take it to daily practice. The lower MAP was frequently higher than the prespecified target in many studies. Not only does this often prevent clear separation between groups, but also, more importantly, it does not allow us to determine that the prespecified target was safe, as it was not formally tested. D'Amico et al (5) elegantly tackled this issue by grouping according to lower and higher MAP targets, but it is difficult to translate the results at bedside due to lack of a clear definition of lower MAP. Most crucially, we know that blood pressure target studies are challenging, and the original studies that feed this meta-analysis have major limitations. Most studies used a "one size fits all" design. Patients in one group were randomized to a MAP range, whereas the other group was assigned to another MAP target. It is likely that some patients were harmed while others benefitted from the intervention. Patients were often kept at the MAP target without considering the results of the intervention (vasopressors, fluids, blood products, anesthetic depth, or even sepsis source control) that got them to the target, even if there were no signs of improvement or even deterioration in organ function. Using different indices of organ perfusion, it has been shown that there was a huge individual variability in the response to a MAP challenge (8,9). This typically also draws from the micro-macro-circulation discordance. Accordingly, the intervention was continued in several patients even when proven detrimental. None of the trials considered other factors influencing organ perfusion. Increasing MAP in a patient with heart failure may compromise cardiac output while it would have a lesser impact in patients with normal cardiac function. Also, central venous pressure (CVP) was not considered, thereby negating any effects of perfusion pressure. However, a given MAP results in a lower perfusion pressure in a patient with high CVP than in a patient with low CVP. This factor may be crucial for some organs like the kidney, where a low renal perfusion pressure has been associated with acute kidney injury (10). Additionally, the usual (or true baseline) MAP of the patient is often not considered when defining the target MAP. Patients with low premorbid MAP are often exposed to higher doses of vasopressors and this may result in a prolonged ICU stay (11). Randomizing patients with different premorbid MAP to the same predefined MAP targets obviously results in variable outcomes and noisy results. Finally, none of the trials used other indices of poor perfusion as entry criteria. Patients suffered from a given condition and then were randomized to a given MAP target. Patients with preserved indices of tissue perfusion may tolerate hypotension better (12). Furthermore, submitting patients with normalized tissue perfusion to interventions such as fluid or MAP challenges increased mortality (13). WHERE DO WE GO FROM HERE? Some may argue that the absence of benefit of systematically increasing MAP suggests that the association of hypotension with organ dysfunction and death is just reflecting the severity of the disease but is not part of the pathophysiological mechanism of organ dysfunction. In other words, it is a simple association rather than causation, and we should not focus on attempting to correct it. However, one should not neglect the potential adverse effects of intervention aiming at increasing MAP. This includes an increased risk of many such outcomes, including but not limited to atrial fibrillation, metabolic effects, and immunosuppression to name just a few. Accordingly, it seems logical from these data to individualize MAP based on different possibilities (Table 1). First, the MAP target should be guided by indices of organ perfusion. We should develop better indices of poor tolerance to hypotension. We nowadays use capillary refill time, lactate, urine output, and mental state in awake patients to assess tolerance to hypotension. Admittedly, these lack specificity and often take some time to develop. The ideal marker should be continuous and noninvasive, assessing directly based on the lines of the impact of perfusion pressure in a relevant organ. TABLE 1. - Future Directions for Evidence Generation on Blood Pressure Targets in Critical Care 1. Individualize MAP targets guided by indices of organ perfusion 2. A specific, continuous, and noninvasive marker of tolerance to blood pressure 3. Consider nocturnal and asleep MAP as a target under sedation and anesthesia 4. Comorbid disease states and cardiac function should be factored in the decision for MAP targets 5. Integrate response (including effectiveness and tolerance) to therapy used to increase MAP MAP = mean arterial pressure. Second, the usual MAP should be considered. However, the usual MAP in an awake active patient may still be too high when sedated. For patients submitted to anesthetic procedures, nocturnal MAP may be considered. Anesthesia, as sleep, decreases oxygen requirements and MAP. During sleep, MAP is usually lower than during daytime. Measuring MAP during sleep and targeting this level during anesthesia (coupled with indices of perfusion) sounds attractive. In the future, measuring MAP and tissue perfusion before surgery with a continuous wearable and noninvasive monitoring device would allow us to better define the MAP to target the different phases of anesthesia/surgery and postsurgical care. Continued in the ICU, may also prevent useless interventions as well as expectations to keep MAP at a certain target for no known reason. In addition, underlying comorbid conditions should be considered to adapt MAP targets. These may include long-standing hypertension, heart failure, significant carotid, or renal artery disease/stenosis, and or a combination. Finally, targeting MAP should integrate the response to therapy, based on effectiveness (organ perfusion) and tolerance (atrial fibrillation). This approach is in line with the current guidelines for sepsis management (14). However, the personalization of MAP targets remains one of the top research priorities identified by the Surviving Sepsis Campaign Research Committee (15). In conclusion, finding the optimal MAP target in critically ill and perioperative patients remains challenging. Subjecting all patients to a higher rather than a lower MAP target is not desired and can even be dangerous, as nicely demonstrated by D'Amico et al (5). However, the heterogeneity in the design of the included trials, difficulty in interpreting MAP targets, and the type of patients included do not allow for the identification of relevant subgroups where a higher MAP target could be beneficial. Individualizing MAP targets is the ideal way to go but several barriers still need to be overcome. Until such time, we suggest using the best clinical judgment to choose blood pressure goals and not be bound to hard targets.